N-channel JFET as Voltage controlled resistor gives unwanted Distorsion

Thread Starter

jacopo1919

Joined Apr 12, 2020
112
Hello! I'm trying to replace the circuitry of a 606 Roland kickdrum that has a pair of resistors tied to ground that determine the Pitch.
606k.jpgI want to Voltage control that resistance. To do so, I can’t use OTA (as LM13700) as single ended resistors ( I wouldn't be able to reach very low resistance values).

I’m trying N-channel JFET (using a J109 at the moment. it has a very low Rds of 12 Ohm).
I'm tweaking with a part of Ray Wilson's schematic. the VCr section is this:

Schermata 2021-11-22 alle 22.46.07.png
The problem i face is that the negative half of the resulting waveform is heavily distorted.
 photo_2021-11-22_22-03-50 copy.jpg
From what i 
I read, Mosfet as VCR should be used mainly for Control Voltage and not in audio Path because they introduce harmonic distorsion but I don’t think this is the real issue in this case.



What am I doing wrong?
 

Papabravo

Joined Feb 24, 2006
21,157
As I see it, the FET provides a resistance that is NOT a linear function of the of Vgs (the gate to source voltage). since you have one end of the FET connected to the voltage signal, when it goes negative, that decreases the Vgs, and if the voltage on the gate cannot go very much lower than the source you are effectively "pinching off" the FET channel limiting the current that goes to GND. I'm curious why you think this would work; JFETs are NOT bipolar devices.

Q: Are your opamps powered by ±12V, or is this a single supply design?
 

Audioguru again

Joined Oct 21, 2019
6,671
A Jfet uses as a voltage controlled resistor must have its source pin common (grounded) and have its drain pin with no more than 50mV of signal. Distortion is reduced by feeding half the signal level at the grain to the gate.
Your Jfet is connected backwards. Look here:
 

Attachments

Papabravo

Joined Feb 24, 2006
21,157
Here is the default JFET characteristic in LTspice. Real parts may very. Almost 30KΩ at Vgs=0 volts. The upper curve is definitely NOT linear for Vgs below GROUND. Picking a PN4391 as the JFET, I'll simulate AGa's circuit for you. give me a few minutes -- it's dinner time.

1637621468547.png
 
Last edited:

Thread Starter

jacopo1919

Joined Apr 12, 2020
112
@Papabravo the opamps are powered with +-12V.
I start to understand a bit but I'm still trying to get this concept and I'm still quite confused.

Assuming that the S is connected to Ground (not as in the schematic above), Vgs remains constantif i don't touch any potentiometer. What it changes is Vds when the kickdrum is generated.
so..Everytime VDS goes more negative, it gets closer to VGS (did you mean this by writing "decrease"?). and so, everytime Vds get close to the Pinchoff Voltage, it limits ''gradually'' the current through the JFET channel and it close the channel when Vds > Vgs(off).
This create the distorsion on the negative half. If we were using a Pchan Jfet in the same configuartion, it would cause distorsion on the positive half instead.

Is this correct or I am lost?

@Audioguru again i saw from the original schematic that the Drain and source were upside down, therefore i turned them around (S tied to Ground). I tried also to set them as for original schematic and it works anyway though.
I tried to feed Half input signal to the Gate as for ESP circuitry + reduced the signal but i still get the distorsion.

Thanks for your help
 

Papabravo

Joined Feb 24, 2006
21,157
I'm not sure if I get what is happening, but here it is:

1637625849663.png

The output looks pretty much like the input for the sum of three harmonically related sine waves. the control voltages at 0, -0.25 & -0.75 provide a sligh change in the output peak level.
 

DickCappels

Joined Aug 21, 2008
10,153
If you truly want to use an OTA as a ground-referenced voltage variable resistance just ground the non-inverting input and put a feedback resistor between the output and the inverting input. The resistance seen looking into the summing node will be
Rin = Rf/(G+1) Where:
Rf is the feedback resistance
G is the open loop voltage gain of the OTA.

I would add some small resistance in series with the signal you want to control to reduce problems like ringing and oscillation.
 

LvW

Joined Jun 13, 2013
1,752
If you truly want to use an OTA as a ground-referenced voltage variable resistance just ground the non-inverting input and put a feedback resistor between the output and the inverting input. The resistance seen looking into the summing node will be
Rin = Rf/(G+1) Where:
Rf is the feedback resistance
G is the open loop voltage gain of the OTA.

I would add some small resistance in series with the signal you want to control to reduce problems like ringing and oscillation.
I cannot confirm the given expression for Rin.
* At first, an OTA has no "open loop voltage gain". Instead, the open-loop characteristic of the OTA is given as transconductance gm (unit A/V).
* Secondly, for Rf=0 the expression would reduce to Rin=0, which is the wrong result.
* Assuming an ideal OTA (infinite input and output impedances), the input resistance Rin does not depend on the feedback resistor (which can be selected as Rf=0): Rin=1/gm .
The transconductance gm can be adjusted (varied) with the externally injected bias current Iabc.
 
Last edited:

Thread Starter

jacopo1919

Joined Apr 12, 2020
112
@Papabravo there is no negative distorsion in your simulation.
Is the 3rd graph the 3 different overlapped results of Vout, based on the 3 different Vcntrlv (of the 3 colors)?
In this case that would mean that the PN4391 is not reducing the resistance in its DS channel, since the 3 output waveforms have the almost same amplitude?

@crutschow i want to track both of them with the same Voltage control, but one of the 2 would be biased somehow in order to give a constant offset pitch.

@DickCappels do you mean something like this?
Schermata 2021-11-23 alle 15.20.33.png
and at the node is seen the resistance to ground?

@LvW: are you referring to the above drawing but without RF?

If we take the LM13700 as an example that has a gm ≈ 19.2 x Iabc,
In this case the Rin seen at the node would be:
for Iabc=0.001A(1mA) --> 1kΩ
for Iabc=0.000001A(1mA) -->1MΩ

that OTA absolute maximum Iabc is 2mA--> 500Ω

In this case there is a minimum resistance to count. Not perfectly what i'm running after but quite good.
Anyway, since this last concept looks similar to the single ended resistor Example in fig.27 of Lm13700 datasheet (Picture below),
Schermata 2021-11-23 alle 15.44.35.png

I believe that the output could have some distorsion without the help of Linearizing Diodes or other tricks unkown to me.

I wanted to use try alternatives to the Ota solution since i have read about MOSFET and JFET methods that could have lead to resistance close to 0Ω.

Thanks a million for all the time you spend to help me.
 

Papabravo

Joined Feb 24, 2006
21,157
Here is the transfer characteristic for the PN4391. It looks much more useful than the default LTspice model. No surprise.

1637627485769.png
@Papabravo there is no negative distorsion in your simulation.
Is the 3rd graph the 3 different overlapped results of Vout, based on the 3 different Vcntrlv (of the 3 colors)?
In this case that would mean that the PN4391 is not reducing the resistance in its DS channel, since the 3 output waveforms have the almost same amplitude?

@crutschow i want to track both of them with the same Voltage control, but one of the 2 would be biased somehow in order to give a constant offset pitch.

@DickCappels do you mean something like this?
View attachment 253317
and at the node is seen the resistance to ground?

@LvW: are you referring to the above drawing but without RF?

If we take the LM13700 as an example that has a gm ≈ 19.2 x Iabc,
In this case the Rin seen at the node would be:
for Iabc=0.001A(1mA) --> 1kΩ
for Iabc=0.000001A(1mA) -->1MΩ

that OTA absolute maximum Iabc is 2mA--> 500Ω

In this case there is a minimum resistance to count. Not perfectly what i'm running after but quite good.
Anyway, since this last concept looks similar to the single ended resistor Example in fig.27 of Lm13700 datasheet (Picture below),
View attachment 253319

I believe that the output could have some distorsion without the help of Linearizing Diodes or other tricks unkown to me.

I wanted to use try alternatives to the Ota solution since i have read about MOSFET and JFET methods that could have lead to resistance close to 0Ω.

Thanks a million for all the time you spend to help me.
Yes, the control voltage produces slight chnages in gain at the output and an FFT confirms there is little or no distortion. You can see faint traces of blue and green color at the peaks. The resistance in the channel is changing, but not as much as you might like. I did simulate the characteristics of the PN4391 and it has a linear regions from Vgs=0V to Vgs=-3.5V where the resistance changes from 110 Ω to about 1020 Ω over that range. You never specified the resistance range you were looking for, so excuse me for trying to help.
 

LvW

Joined Jun 13, 2013
1,752
[QUOTE="jacopo1919, post: 1684842, member: 694059"

@LvW: are you referring to the above drawing but without RF?

If we take the LM13700 as an example that has a gm ≈ 19.2 x Iabc,
In this case the Rin seen at the node would be:
for Iabc=0.001A(1mA) --> 1kΩ
for Iabc=0.000001A(1mA) -->1MΩ

[/QUOTE]

With RF= finite or RF=0 (no difference as long as (r_out>>RF).
However - the input resistance is NOT simply 1/Iabc (as in your calculation).
As I wrote: Rin=1/gm.
 

Papabravo

Joined Feb 24, 2006
21,157
I have tried to recreate your source signal and I think it shows that the FET provids offset and changes the gain for low level signals. I don't see any distortion on the negative going peaks.

1637705393058.png
 

ronsimpson

Joined Oct 7, 2019
2,985
When I was making audio processing "things" in 1975-80 I used the feed back resistors R2, R3 in posts #6 & #13. This helped the distortion. Also keeping the signal level down helps. I also used MOSFETs with feed back resistors at a higher level of signal.

Another option is to use a Light Dependent Resistor and a LED. (photo call) The distortion is good but there is a time lag in changing the resistance. I think in your case you will not care about 10mS delay. I used the CAD cell and LED in a black tube.

Another option is a digital POT. This one needs a computer to set its value but some have a UP and DOWN buttons.
1637712992949.png
 

Thread Starter

jacopo1919

Joined Apr 12, 2020
112
@Papabravo thanks for the graph.

It looks like there is no offset for the second waveform which initially sits on 0V. The rest looks quite good (beside the fact that the amplification doesn't look linear , but i believe that this should be the normal behaviour since the Vds, which follows Vin, is changing .. changing the channel resistance as well)

I was doing something wrong because the distorsion were appearing only when i was using a J109 NFET. Now I tried with a NFET J211 and the distorsion stops to appear. ( i need to purchase a PN4391 and test it).

As soon as i get home i will check on the datasheet and see if i can understand the reason of that distorsion.

The range "needed" is 4k to 0ohm for my application but this should depend on the feedback resistor as well.
In the same circuit, a resistance from 48kΩ to around 0Ω would come handy in order to control the decay.
So basically i'm trying to understand the potential/pro/cons of making a variable resistance using FETs instead of an OTA (which is the only method that I tried so far)
 

Papabravo

Joined Feb 24, 2006
21,157
I hope you chan find the characteristic curves you need to estimate the range of resistance of each JFET. Problem is that even devices from tha same wafer may have different characterisatics. It is generally a mistake to base a design on specific device parameters. Since this is probably a one-off you can afford to buy a bucket of JFETs and look for the golden part you seek. Good luck on your quest.
 

DickCappels

Joined Aug 21, 2008
10,153
In my first job in electronics I was at a small manufacturer of medical imaging devices in California. It was necessary to bin and mark the pinch-off range of cheap JFETs that were then used in pairs for gain control of video amplifiers. That's the time and place I was told to use 10% resistors whenever possible because they were cheaper than 5% resistors. The good old days.
 

Thread Starter

jacopo1919

Joined Apr 12, 2020
112
@Papabravo i see the point. For example:
as for 109 datasheet, the pinchoff may vary from -2 V to -6 V. Quite a wide range.

This might be off topic but i wouldn't open a thread only for this:
Do you want to introduce other methods to achieve a voltage controlled resistance, so that i can check them out?

I tried only Vactrols, Ota (the LM13700), JFET and i'm very curious about new methods.

@DickCappels yes i did found that note. Thanks. I read it once but i need to check it again to better understand.
 
Top